UnifyBench

RankingMMLU-Pro

MMLU-Pro

Data updated 22 Oct 2026

Bucket
Knowledge
Unit
percent
Direction
Higher is better
Version
—
Display harness
MMLU-Pro reported
Board
https://github.com/TIGER-AI-Lab/MMLU-Pro

Compare published benchmark results with category weights →

Models

Results are ordered by score in the display harness, followed by models with no result. Other harnesses are listed separately below. Missing results remain unknown.

1–25 of 471 entries

ModelScoreEvidenceSource-recorded date
Gemini 3.1 ProGoogle91.16%official board2026-10-13
MiniMax-M2.1MiniMax88%official board2026-10-13
Qwen3.5-397B-A17BQwen87.8%official board2026-10-13
GPT-5.4OpenAI87.5%official board2026-10-13
GPT-5.2OpenAI87.4%official board2026-10-13
NVIDIA Nemotron 3 UltraNVIDIA86.8%lab self-report

Current source benchmark table does not expose this fixture value or its task/window configuration; recover a pinned historical report before admitting.

2026-06-04
Qwen3.5-122B-A10BQwen86.7%official board2026-10-13
GPT-5.1OpenAI86.4%official board2026-10-13
Qwen3.5-27BQwen86.1%official board2026-10-13
Gemini 2.5 ProGoogle86%official board2026-10-13
GLM-5Z.ai86%official board2026-10-13
Qwen3.5-35B-A3BQwen85.3%official board2026-10-13
GLM-4.5Z.ai84.6%official board2026-10-13
Qwen3-235B-A22B-Thinking-2507Qwen84.5%official board2026-10-13
DeepSeek-R1-0528DeepSeek83.4%official board2026-10-13
Qwen3-235B-A22B-Instruct-2507Qwen83%official board2026-10-13
Seed OSS 36B InstructByteDance82.7%official board2026-10-13
Qwen3.5-9BQwen82.5%official board2026-10-13
MiniMax-M2MiniMax82%official board2026-10-13
GPT-4.1OpenAI81.8%official board2026-10-13
GLM-4.5-AirZ.ai81.4%official board2026-10-13
Qwen3-30B-A3B-Thinking-2507Qwen80.9%official board2026-10-13
Llama 4 MaverickMeta80.5%official board2026-10-13
MiniMax-M2.5MiniMax80.1%official board2026-10-13
Qwen3.5-4BQwen79.1%official board2026-10-13